6. LIGHT SCATTERING 6.1 The first Born approximation

In many situations, light interacts with inhomogeneous systems, in which case

the generic light-matter interaction process is referred to as scattering

x

ks

Pr r '

r '

z

y

ki

Figure 6-1. Light scattering by an inhomogeneous medium.

1

The goal in light scattering experiments is to infer information about the

refractive index distribution ( )n r from measurements on the scattered light

We show that this problem can be solved if we assume weakly scattering media.

Recall the Helmholtz equation for a loss-free medium

( ) ( ) ( )( ) ( )

2 2

0

0

, , , 0

, ,

U U

n k

k c

+ =

=

=

r r rr r (6.1)

Equation 1a can be re-written to show the source term on the right hand side,

such that the scattered field satisfies

( ) ( ) ( ) ( )

( ) ( )

2 20

2 20

, , 4 , ,1, , 1

4

U k U F U

F k n

+ =

=

r r r r

r r(6.2)

( ),F r is called the scattering potential associated with the medium. 2

The elementary equation that gives Greens function ( ),h r is

( ) ( ) ( ) ( )22 20, ,h k h + = r r r (6.3)

The solution in the spatial frequency domain is

( ) 2 20

1,hk k

=

k (6.4)

Taking the Fourier transform of Eq. 4, we arrive at the spherical wave solution

( )0

, .ik rer

=h r (6.5)

The solution for the scattered field is a convolution between the source term and

the Green function, ( ),h r ,

( ) ( ) ( )0 '

3, ', ', ''

ikeU F U d

=

r r

r r r rr r

(6.6)

3

The integral in Eq. 6 can be simplified if we assume that the measurements are

performed in the far-zone, 'r r . Thus, we can write 2 2

0

' ' 2 '

'

's

r r

rr

rk

= +

r r rrr r

k r

(6.7)

'r r is the scalar product of vectors r and 'r and 0

sk

k is the unit vector

associated with the direction of propagation.

With this far-zone approximation, Eq. 6 can be re-written as

( ) ( ) ( )0

' 3, ', ', 'sik r

ieU F U e dr

= k rr r r r (6.8)

4

Equation 8 indicates that, far from the scattering medium, the field behaves as a

spherical wave, 0ik re

r, perturbed by the scattering amplitude, defined as

( ) ( ) ( ) ' 3, ', ', 'sisf F U e d = k rk r r r (6.9) To obtain a tractable expression for this integral, assume the scattering is weak,

which allows us to expand ( )',U r . The Born approximation assumes that the

field inside the scattering volume is constant and equal to the incident field,

( ) '', .iiiU e k rr (6.10)

With this approximation, we finally obtain for the scattered field

( ) ( ) ( ) ' 3, ', 's iisf F e d r = k k rk r (6.11) The right hand side is a 3D Fourier transform.

5

Within the first Born approximation, measurements of the field scattered at a

given angle, gives access to the Fourier component s i= q k k of the scattering

potential F,

( ) ( ) ' 3, ', '.if F e d = qrq r r (6.12) q is the difference between the scattered and initial wavevectors, called

scattering wavevector and in quantum mechanics is referred to as the momentum

transfer.

k

ki

q

Figure 2. The momentum transfer.

6

From Fig. 2 it can be seen that 02 sin 2q k= , the scattering angle. Due to the

Fourier integral, Eq. 12 is easily invertible to provide the scattering potential,

( ) ( ) ' 3', , .iF U e d qrr q q (6.13) Equation 13 establishes the solution to the inverse scattering problem, it

provides a way of retrieving information about the medium under investigation

via angular scattering measurements. Measuring U at many scattering angles

allows the reconstruction of F from its Fourier components.

For far-zone measurements at a fixed distance R, the scattering amplitude and

the scattered field differ only by a constant ikRe

R , can be used interchangeably.

7

To retrieve ( )',F r experimentally, two essential conditions must be met:

o The measurement has to provide the complex scattered field o The scattered field has to be measured over an infinite range of spatial

frequencies q (the limits of the integral in Eq. 13 are to )

Finally, although good progress has been made recently in terms of measuring

phase information, most measurements are intensity-based. Note that if one has

access only to the intensity of the scattered light, ( ) 2,U q , then the

autocorrelation of ( )',F r is retrieved, and not F itself,

( ) ( ) ( )2 ' 3, ', ',iU q e d F r F r = qr q (6.14) The result in Eq. 14 is simply the correlation theorem applied to 3D Fourier

transforms.

8

Secondly, we have experimental access to limited frequency range, or bandwidth.

The spatial frequency (or momentum transfer) range is intrinsically limited. Thus, for

a given incident wave vector ki, with 0ik k= , the highest possible q is obtained for

backscattering, 02b i k =k k (Fig. 3a)

qb

kb kix

a) b)

qb

kb ki Figure 6-3. Momentum transfer for backscattering configuration for ||i xk and ||i xk .

Similarly for an incident wave vector in the opposite direction, ik , the maximum

momentum transfer is also 02q k= (Fig. 3b). Altogether, the maximum frequency

coverage is 04k , as illustrated in Fig. 4.

9

kb ki ki kb

2k02k0 Figure 6-4. Ewald scattering sphere.

As we rotate the incident wavevector from ik to ik , the backscattering wavevector

rotates from bk to bk , such that the tip of q describes a sphere of radius 02k . This is

known as the Ewald sphere, or Ewald limiting sphere.

10

Let us study the effect of this bandwidth limitation in the best case scenario of

Ewalds sphere coverage. The measured (truncated field), ( ),U q , can be expressed

as

( ) ( ) 0, ,0 2

,0, restU q k

U

=q

q (6.15)

Using the definition of a rectangular function in 3D, the ball function, we can

express U as

( ) ( )0

, ,4qU Uk

=

q q (6.16)

Thus, the scattering potential retrieved by measuring the scattered field U can be

written as (from Eq. 13)

( ) ( ) ( )', ', 'F F r = r r (6.17)

11

In Eq. 17, is the Fourier transform of the ball function 04

qk

, which has the

form

( ) ( ) ( )( )

( )30 0 0 030

sin 2 ' 2 'cos 2 '' 2

2k r k r k r

r kk r

= (6.18)

It follows that even in the best case scenario, i.e. full Ewald sphere coverage, the

reconstructed object ( )',F r is a smooth version of the original object, where the

smoothing function is ( )'r . Practically, to cover the entire Ewald sphere, requires

illuminating the object from all directions and measure the scattered complex field

over the entire solid angle, for each illumination. This is, of course, a challenging

task, rarely achieved in practice. Instead, fewer measurements are performed, at the

expense of degrading resolution.

Figure 5 depicts the 1D profile of the 3D function ( )'r . 12

11.45

12

r '( )

1.452k0r '

Figure 6-5. Profile through function ( )'r in Eq. 18.

The FWHM of this function, 'r , is given by 02 ' 2k r q = , or ' 0.23r = . We can,

therefore, conclude that the best achievable resolution in reconstructing the 3D

object is approximately / 4 , only a factor of 2 below the common half-wavelength

limit. This factor of 2 gain is dues to illumination from both sides.

13

6.2 Scattering cross sections of single particles. Here by particle, we mean a region in space that is characterized by a dielectric

permeability 2n = , which is different from that of the surrounding medium

(Fig. 6).

U0

ki

Us r( )r

ks

Figure 6-6. Light scattering by a single particle.

14

The field scattered in the far zone has the general form of a perturbed spherical

wave,

( ) ( )0 , ,ikr

s s ie fr

= U r U k k (6.19)

where ( )sU r is the scattered field at position r, r = r , and ( ),s if k k defines the

scattering amplitude. The function f is physically similar to that encountered above,

when discussing the Born approximation, (Eq. 9) except that here it also includes

polarization in formation, in addition to the amplitude and phase.

The differential cross-section associated with the particle is defined as

( ) 2, lim ,sd s i ri

r

=Sk kS

(6.20)

where sS and iS are the Poynting vectors associated along the scattered and initial

direction, respectively, with moduli 2

, ,1

2i s i sU

=S .

15

Note that, according to Eq. 20, d is defined in the far-zone. It follows immediately

that the differential cross-section equals the modulus squared of the scattering

amplitude,

( ) ( ) 2, , .d s i s if =k k k k (6.21)

Note that the unit for d is [ ] 2 /d m rad = . One particular case is obtained for backscattering, when s i= k k ,

( ), ,b d i i = k k (6.22)

where b is referred to as the backscattering cross section.

The normalized version of d defines the so-called phase function,

( ) ( )( )

4

,, 4 .

,d s i

s id s i

pd

=

k k

k kk k

(6.23)

16

The phase function p defines the angular probability density associated with the

scattered light. The integral in the denominator of Eq. 22 defines the scattering cross

section.

( )4

,s d s i d

= k k (6.24)

Thus, the unit of the scattering cross section is [ ] 2s m = . In general, if the particle also absorbs light, we can define an analogous absorption cross section, such that the

attenuation due to the combined effect is governed by a total cross section,

a s = + (6.25)

For particles of arbitrary shapes, sizes, and refractive indices, deriving expression for

the scattering cross sections, d and s , is very difficult. However, if simplifying

assumptions can be made, the problem becomes tractable. In the following, we will

investigate some of these particular cases, most commonly encountered in practice.

17

6.3 Particles under the Born approximation When the refractive index of a particle is only slightly different from that of the

surrounding medium, its scattering properties can be derived analytically within the

framework of the Born approximation. Thus the scattered amplitude from such a

particle is the Fourier transform of the scattering potential of the particle (Eq. 12)

( ) ( ) ' 3, ' ',ipf F e d = qrq r r (6.26) In Eq. 26, pF is the scattering potential of the particle. We can conclude that the

scattering amplitude and the scattering potential form a Fourier pair, up to a constant

0k

( ) ( )0

1, ',f Fk

q r (6.27)

18

The rule of thumb for this scattering regime to apply is that the total phase shift

accumulation through the particle is small, say, 1 rad< . For a particle of diameter d

and refractive index n in air, this condition is ( ) 01 1n k d < . Under these conditions,

the problem becomes easily tractable for arbitrarily shaped particles. For intricate

shapes, the 3D Fourier transform in Eq. 26 can be at least solved numerically. For

some regular shapes, we can find the scattered field in analytic form, as described

below.

6.3.1 Spherical particles

For a spherical particle, the scattering potential has the form of the ball function

( )

( )

0

2 20 0

''2

1 14

prF Fr

F k n

=

=

r (6.28)

19

Equations 28a-b establishes that the particle is spherical in shape and is characterized

by a constant scattering potential 0F inside (and zero outside)

Thus, plugging Eqs. 28 into Eq. 26, we obtain the scattering amplitude distribution,

( ) ( ) ( ) ( )( )

2 2 30 3

sin cos, 1

qr qr qrf n k r

qr

q (6.29)

The differential cross section is

( ) ( )

( ) ( ) ( )( )

2

222 2 4

0 3

, ,

sin cos1

d f

qr qr qrn V k

qr

=

q q

(6.30)

where 34 / 3V r= is the volume of the particle.

Figure 7 illustrates the angular scattering according to Eq. 28.

20

d ( )

Figure 6-7. Angular scattering for a Rayleigh-Gauss particle with sin 2qr= .

Equation 30 establishes the differential cross section associated with a spherical

particle under the Born approximation. Sometimes this scattering regime is referred

to as Rayleigh-Gauss, and the particle for which this formula holds as Rayleigh-

21

Gauss particles. The scattering angle enters explicitly Eq. 28, by expressing

02 sin 2q k= .

A very interesting particular case is obtained when 0qr . Note that this

asymptotical case may happen when the particle is very small, but also when the

measurement is performed at very small angles, i.e. 0 . If we expand around the

origin the function,

( )2

3

23 2

3

sin cos

16 2

16

x x xf xx

x xx x

x

=

=

(6.31)

22

Thus, measurements at small scattering angles, can reveal volume of the particle,

( ) ( )22 2 400, 1d q n V k q (6.32) This result is the basis for many flow cytometry instruments, where the volume of

cells is estimated via measurements of forward scattering. The cell structure, i.e.

higher frequency components are retrieved through larger angle, or side scattering.

On the other hand, Eq. 31 applies equally well where the particle is very small,

i.e. Rayleigh particle. In this case, d is independent of angles, i.e. Rayleigh

scattering is isotropic. Thus, the scattering cross section for Rayleigh particles has

the form

( ) ( )2 4 201 .s n k V (6.33)

The scattering cross section of Rayleigh particles is characterized by strong

dependence on size and wavelength

23

64s

s

r

(6.34)

The nanoparticles in the atmosphere have scattering cross sections that are 16 times

larger for a wavelength 400b nm = (blue) than for 800r nm = (red). This explains

why the clear sky looks bluish, and the sun itself looks reddish.

6.3.2 Cubical particles

For a cubical particle, the scattering potential has the form

( )

( )

0

2 20 0

, ,2 2 2

1 14

x y zF x y z Fa b c

F k n

=

=

(6.35)

The scattering amplitude in this case is

( ) ( ) ( ) ( ) ( )2 20, , 1 sinc sinc sincx y z x y zf q q q n k V q a q b q c= (6.36)

24

qki

ks

z

x

c

a b y

Figure 6-8. Scattering by a cubical particle.

It can be seen that the differential cross section 2d f = has the same 4V and 40k

dependence as for spherical particles.

6.3.3 Cylindrical particles

For a cylindrical particle, the scattering potential is

25

( )

( )

2 2

0

220

0

, ,2 2

14

x y zF x y z Fa b

kF k

+ =

=

(6.37)

The 3D Fourier transform of F yields the scattering amplitude

( ) ( )2 2

12

2 2, , sinc

2

x y

x y z z

x y

J q q af q q q a b q b

q q

+ =

+ (6.38)

26

ab

z

x

y

b2

Figure 6-9. Scattering by a cylindrical particle.

As before, d and s can be easily obtained from 2f .

27

6.4 Scattering from ensembles of particles under the Born approximation. Let us consider the situation where the scattering experiment is performed over an

ensemble of particles randomly distributed in space, as illustrated in Fig. 10.

ks

r

y

z

x

ki

Figure 6-10. Scattering by an ensemble of particles.

If we assume that the ensemble is made of identical particles of scattering potential

( )0F r , then the ensemble scattering potential can be expressed via -functions as 28

( ) ( ) ( )0 jj

F F = r r r r (6.39)

Equation 39 establishes the discrete distribution of the scattering potential, where

each particle is positioned at position ir . Note that, in order for the Born

approximation to apply, the particle distribution is space, which eliminates the

possibility of overlapping particles. The scattering amplitude is simply the 3D

Fourier transform of ( )F r ,

( ) ( ) ( )( )

30

0 ,j

ij

i

f F e d

f e

=

=

qr

qr

q r r r r

q (6.40)

where, as before, s i= q k k .

Therefore, the scattering amplitude of the ensemble is the scattering amplitude

of a single particle, ( )0f q , multiplied by the so-called structure function, defined as

29

( ) jij

S q e= qk (6.41)

We can express the scattering amplitude as a product,

( ) ( ) ( )0f f S q= q q (6.42)

In Eq. 42, ( )0f q is sometimes referred to as the form function. The physical meaning

of the form and structure functions becomes apparent if we note that the size of the

particle is smaller than the inter-particle distance. It follows that ( )0f q is a broader

function than ( )S q , or that ( )50f acts as the envelope (form) of ( )f q and ( )S q as its

rapidly varying component (structure).

Example Scattering from 2 spherical particles (radius a) separated by a distance (b)

(Fig. 11).

30

zb2

x

y

Figure 6-11. Scattering by two particles separated by a distance b.

According to Eq. 42, the far-zone scattering amplitude is

( ) ( ) ( )( ) ( ) ( )

( )( )

0

2 20 3

cos 2sin cos

1 cos 2

z

z

bf f q

qr qr qr bn k V qqr

=

=

q q (6.43)

31

In Eq. 43 we wrote explicitly the form factor of a Rayleigh-Gauss particle of volume

V and refractive index n.

This approach is the basis for extracting crystal structures form x-ray scattering

measurements.

32

6.5. Mie scattering. Mie provided the full electromagnetic solutions of Maxwells equations for a

spherical particle of arbitrary size and refractive index. The scattering cross section

has the form

( )( )2 22 21

2 2 1 ,s n nn

a n a b

=

= + + (6.44)

where an and bn are complicated functions of 0k k a= , 0 0/k na n = , with a the radius

of the particle, 0k the wavenumber in the medium, 0n the refractive index of the

medium, and n the refractive index of the particle. Equation 44 shows that the Mie

solution is in terms of an infinite series, which can only be evaluated numerically.

Although today common personal computers can evaluate s very fast, we should

note that as the particle increases in size, the summation converges more slowly, as a

33

higher number of terms contribute significantly. Physically, as a increases, standing

waves (nodes) with higher number of maxima and minima fit inside the sphere.

Although restricted to spherical particles, Mie theory is heavily used for tissue

modeling.

34

6.6. Multiple Scattering Regime In the previous section, we introduced the basic description of the single

scattering process, which applies to sparse distributions of particles. For the

single scattering regime, the average dimension of the scattering medium must

be smaller than the mean free path ls.

In many situations this is not the case and the light travels through the medium

over distances much longer than ls and encounters many scattering events during

the propagation. For this case, the simple treatment presented earlier does not

apply and the complexity of the problem is increased.

A full vector solution to the field propagation in dense media is not available;

scalar theories give results in an analytical form only for special, intermediate

cases of multiple scattering regimes.

35

Figure 12. Scattering medium under transport conditions.

Simplifications can be used when the scattering medium is very dense and the

electromagnetic waves can be assumed to travel randomly through the medium.

36

6.6.1. Elements of Radiative Transport Theory

The transport description in the case of a highly scattering medium is based on a

photon random walk picture similar to that used in solid state physics and

nuclear reactor physics [7]. The wave correlations that may survive this heavy

scattering regime are neglected, while each scattering event is considered an

independent process.

Using the energy (photon) conservation principle, the final goal of the transport

theory is to find the time-dependent photon distribution at each point in the

scattering medium. We will make use of the following definitions and notations:

o N(r,t) photon density [m3]; o v s = v/ls frequency of interaction [s-1]; o = a + s attenuation coefficient [m]; o (r, t) = vN(r, t) photon flux [m2 s1]; o n(r,,t) density of photons (m3sr1) having the direction of propagation

37

within the solid angle interval [, + d];

o (r, , t)R= v n(r, , t) angular photon flux [m2s1sr1]; scalar o j(r, , t) = (r, , t) angular current density [m2s1sr1]; vector o J(r, t) =

4

( , , )j t d

r photon current density [m2s1].

To write the transport equation, we first identify all the contributions that affect

the photon balance within a specific volume V. We obtain the following

transport equation in its integral form:

[ ] 34

( , , ) ' ( ' ) ( , ', ) '0.

( , , )

sV

dn v n v n t v n t ddt d rds t

+ + =

r r

r (6.45)

The first term denotes the rate of change in the energy density, the second term

represents the total leakage out of the boundary, the third term is the loss due to

38

scattering, the fourth term accounts for the photons that are scattered into the

direction , while the last term represents the possible source of photons inside

the volume.

Since the volume V was arbitrarily chosen, the integral of Eq. 2.31 is zero if and

only if the integrand vanishes. Thus, the differential form of the transport

equation is:

[ ]4

( , , )

' ( ' ) ( , ', ) ' ( , , ) 0.s

dn v n v n tdt

v n r t d s t

+ +

=

r

r (6.46)

39

The same equation can be expressed in terms of the angular flux (r, , t) = v n(r, , t):

[ ]4

1 ( , , )

( ' ) ( , ', ) ' ( , , ) 0.s

d tv dt

t d s t

+ +

=

r

r r (6.47)

The initial condition for this equation is (r, , 0) = 0(r, ), while the

boundary condition is (rs,,t)rsS = 0, if es < 0. The transport equation can

be solved in most cases only numerically. As we will see in the next section,

further simplifications to this equation are capable of providing solutions in

closed forms and will, therefore, receive a special attention.

40

6.6.2. The Diffusion Approximation of the Transport Theory

Eq. 2.33 can be integrated with respect to the solid angle , the result can be

expressed in terms of the photon flux and the current density J as a continuity

equation: 1 ( , ) ( , ) ( , ) ( , )s

d t t t s tv dt

+ + = +J r r r r (6.48)

Eq. 2.34 contains two independent variables and J:

4

4

( , ) ( , , )

( , ) ( , , )

t t d

J t t d

=

=

r r

r r (6.49)

An additional assumption is needed to make the computations tractable. In most

cases, one can assume that the angular flux is only linearly anisotropic: 1 3( , , ) ( , ) ( , ) .

4 4t t J t

+ r r r (6.50)

41

This represents the first order Legendre expansion of the angular flux and is the

approximation that lets us obtain the diffusion equation. It is called the P1

approximation, since it represents a Legendre polynomial Pl, for l = 1.

The photon source is considered to be isotropic and the photon current to vary

slowly with time: 1 tvt

phase function defined in Eq. 2.6 as the weighting function. g accounts for the

fact that, for large particles, the scattering pattern is not isotropic.

In Eq. 2.37, D(r) is the diffusion coefficient of photons in the medium and is the

only parameter that contains information about the scattering medium.

Combining Eqs. 2.34 and 2.37, we can obtain the so-called diffusion equation

[ ]1 ( ) ( , ) ( ) ( , ) ( , ).aD r t t S tv t

+ =

r r r r (6.53)

a is the absorption coefficient and S(r,t) is the photon source considered to be isotropic.

The diffusion equation breaks down for short times of evolution, t < lt/c. It takes

a finite amount of time for the diffusion regime to establish.

Although restricted to a particular situation of multiple light scattering, the

diffusion treatment finds important applications in practice.

43

Using appropriate boundary conditions, Eq. 2.39 can be solved for particular

geometries. Once the photon flux is calculated, the current density J can be

obtained using Ficks law [7].

The current density J(r,t) is the measurable quantity in a scattering experiment

and thus deserves special attention. The total optical power measured by a

detector is proportional with the integral of J(r,t) over the area of detection. t

represents the time of flight for a photon between the moment of emission and

that of detection. When a constant group velocity is assumed, this variable can

be transformed into an equivalent one- the optical path-length, s = vt.

The path-length probability density0

( ) ( ) / ( )P s J s J s ds

= , the probability that

the photons traveled, within the sample an equivalent optical path-length in the

interval (s, s + ds), can also be evaluated.

44